A Triumvirate of AI Driven Theoretical Discovery
arXiv:2405.19973 · doi:10.1038/s42254-024-00740-1
Abstract
Recent years have seen the dramatic rise of the usage of AI algorithms in pure mathematics and fundamental sciences such as theoretical physics. This is perhaps counter-intuitive since mathematical sciences require the rigorous definitions, derivations, and proofs, in contrast to the experimental sciences which rely on the modelling of data with error-bars. In this Perspective, we categorize the approaches to mathematical discovery as "top-down", "bottom-up" and "meta-mathematics", as inspired by historical examples. We review some of the progress over the last few years, comparing and contrasting both the advances and the short-comings in each approach. We argue that while the theorist is in no way in danger of being replaced by AI in the near future, the hybrid of human expertise and AI algorithms will become an integral part of theoretical discovery.
arXiv comment14 pages, under consideration for Nature Review Physics
References in corpus (40)
- LLaMA: Open and Efficient Foundation Language Models
- Machine Learning of Calabi-Yau Volumes
- Machine Learning in the String Landscape
- Deep Learning and AdS/CFT
- Evolving neural networks with genetic algorithms to study the String Landscape
- Machine Learning CICY Threefolds
- Hilbert's Program Then and Now
- Machine Learning and Computational Mathematics
- Machine Learning Line Bundle Cohomologies of Hypersurfaces in Toric Varieties
- (Machine) Learning to Do More with Less
- Machine learning Calabi-Yau metrics
- The Ramanujan Machine: Automatically Generated Conjectures on Fundamental Constants
- Deep learning in the heterotic orbifold landscape
- Quiver Mutations, Seiberg Duality and Machine Learning
- Machine Learning Calabi-Yau Hypersurfaces
- Nonperturbative renormalization for the neural network-QFT correspondence
- Is Deep Learning a Renormalization Group Flow?
- Machine learning for complete intersection Calabi-Yau manifolds: a methodological study
- Scientific intuition inspired by machine learning generated hypotheses
- Fundamental limits to learning closed-form mathematical models from data
- Estimating Calabi-Yau Hypersurface and Triangulation Counts with Equation Learners
- Explore and Exploit with Heterotic Line Bundle Models
- Formulae for Line Bundle Cohomology on Calabi-Yau Threefolds
- Deep learning and k-means clustering in heterotic string vacua with line bundles
- Machine Learning Lie Structures & Applications to Physics
- Machine Learning and Algebraic Approaches towards Complete Matter Spectra in 4d F-theory
- Hilbert Series, Machine Learning, and Applications to Physics
- Artificial Neural Network in Cosmic Landscape
- The landscape of promising non-supersymmetric string models
- Machine Learning for Mathematical Software
- Disentangling a Deep Learned Volume Formula
- Large Language Models for Mathematical Reasoning: Progresses and Challenges
- Learning to Inflate
- Ising Machines for Diophantine Problems in Physics
- Symbolic Equation Solving via Reinforcement Learning
- Machine learning the dimension of a Fano variety
- A Neural Network Perturbation Theory Based on the Born Series
- Metacognitive Capabilities of LLMs: An Exploration in Mathematical Problem Solving
- Why Machines Cannot Learn Mathematics, Yet
- Machine learning of the well known things